The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted JSON data files that could potentially contain malicious natural language instructions designed to influence the agent's behavior during the pre-sales engineering workflow. * Ingestion points: The skill ingests requirements data from 'assets/sample_rfp_data.json' and expects the agent to process other JSON inputs such as 'competitive_data.json' and 'poc_data.json'. * Boundary markers: There are no explicit delimiters or 'ignore embedded instructions' warnings provided in the workflow for the data being processed. * Capability inventory: The skill allows the execution of local Python scripts (rfp_response_analyzer.py, competitive_matrix_builder.py, poc_planner.py) via shell commands. These scripts perform data analysis and validation but do not possess network access, privileged file system access, or other high-risk capabilities. * Sanitization: While the Python scripts use standard JSON parsing, they do not include validation logic to identify or neutralize potential prompt injection patterns within the ingested data.

sales-engineer